Can you spot fake motion blur? (When its next to the real thing)
Вставка
- Опубліковано 15 жов 2024
- Adding 'fake' digital motion blur in post is more and more common these days, but can you spot it next to the real thing?
Here, I side by side three various digital motion blurs with a real 180 degree shutter in camera motion blur.
(Some of these digital blurs taking a longer to render than others.)
To achieve this I filmed with two side by side cameras on my FPV quad, one with a fast 1/1600 shutter speed, then other at 1/50 shutter speed.
Then adding various digital motion blur effects onto the fast shutter speed version.
The chap that did the digital motion blur for me here is-
Hürcan Emre Yılmazer
hurcan.yilmazer on instagram
And the motion blurs he used were-
RSMB is ReelSmart Motion Bur Plugin.
The same plugin in after effects and davinci resolve.
AE_pixel_motion_blur is after effect's own included motion blur effect
And default MB is Davinci Resolve Studio's own included effect.
Any other questions, pop them in the comments.
Link to Hürcan Emre Yılmazer IG-
www.instagram....
Here is the previous vid (real motion blur vs NO motion blur) - ua-cam.com/video/jdYSZ95NG7c/v-deo.html
You'd think there'd be a ton o' video comparisons on real vs 'fake" motion-blur, but noooo. Thank you for creating this! Exactly what I was looking for. Cheers
Glad it was helpful.
One of the most useful content I've seens in the last couple of months regarding to video editing. And was not easy to find it :)
My bet was top right, I thought the fake motion blurs would be the more regular-looking.
My conclusion is the same, crispy shots and post-process is the future. ND filter companies just dropped their tear
This is great! Been thinking about shooting handheld in daylight with faster shutterspeeds and gyrodata to have the stabilisation flexibility in post. If this is a feasible way to add motion blur back in post, I don't see any reason not to. I guess low light scenarios or fast turnaround projects still would require a gimbal though.
Having this same dilemma at the moment (A7 III is really hard to get anything remotely useable handheld even when being careful). It's surprisingly not talked about much.. everyone just seems to blindly suggest gyrodata stablisation without considering some people want a little motion blur. I see a lot of videos that do both and the stabilised motion blur jitters are very distracting.
A top handle seems to reduce shakes a fair bit (or a gorilla pod you modify into a top handle/or hold upside down) and lenses with optical stabilisation are surprisingly good (the 24-105 for example). Both can be quite large though if trying to keep a small profile and there's only a few lenses with optical stabilisation unfortunately.
The A7S III seems to do some magic with it's active stabilisation too where it seems to not get motion blur jitters somehow. Hope we see some more strides in solving this in the future.
I honestly couldn't see any difference until you showed the freeze frames. I guess I'd rather add fake motion blur and avoid baked-in motion blur from stabilization. And imagine if you had software that could use gyro data from camera to avoid having to guess which way to draw the blur. That would be very useful for both stabilization and adding motion blur.
Yep. But bare in mind not all motion blur is from camera movement.
If you were filming a mountain biker. Or a bird flew past etc. The motion blur for other moving objects would be completely different to direction of camera travel.
@@biscuitsalive
When there is someone moving in the scene, in addition to the camera movement, do these fake motion blur techniques still work?
only noticed with pausing the video, then I could tell.
10/10 video. Excellent breakdown.
bottom left, you can easily see it when you pay attention to the corners of the frames, often there are errors in the calculation which is why you can see part of the frame without any motion blur or some weird swirly blur which leads in multiple directions at once.
if i wouldn’t have ever worked with RSMB, i probably wouldn’t have noticed it though, especially when viewing on a mobile screen like i did.
either way, i'd rather spend a few seconds before the flight to attach the right ND than dealing with multiple hours of additional rendertime in post honestly.
edit: you actually explain all that already, should have watched till the end before commenting 😄 sorry!
My thoughts too
Looked for this video for entirely too long 😭
A very helpful video. Thank you. I'm working more and more with my iPhone and I'm finding that using vario ND filters a bit of a pain particularly when using a gimbal. So, I'm looking at using Resolve Studio Motion Blur instead. As an enthusiastic dabbler I suspect that I won't be able to tell the difference between real and post motion blur. Thanks again.
hi, do you know/remember/could figure out which version of the RSMB plugin was used here? just wondering as the video is already a bit older. thanks!
So it is mostly a problem with films that go in a lot of directions? When just moving in one direction it might nog matter that much.
It did take me a moment to zero in on it, but I was ultimately able to tell the bottom left was the real motion blur without freeze framing the video. I have to admit the blurs added in post were more convincing than I'd have expected. I'd be curious to see how the post blurs perform with different shutter angles. Is the result better when there is some interframe motion blur already present (but less than with a 180degree shutter), or when each frame is completely clean? Next test should be stabilization and post motion blur with increasingly fast shutter angles - testing 90 degree, 45, 22.5, etc.
I think it works best when almost no motion blur. The computer needs to identify shapes and certain contrasting elements in each frame so it can determine direction.
Existing “real” motion blur just gives it less information to analyse.
Same for when you are 3D tracking camera moves etc. It works best with no motion blur. This is one of the reasons this is probably the future as post processing is easier for adding effects etc without in camera motion blur.
@@biscuitsalive yeah, that makes sense, I’m used to running up the shutter speed for vfx/green screen/rotoscope stuff. Though, usually anything above 90 degrees has been sufficient. But I suppose since most FPV shots are done in direct sun, there’s little reason not to go to faster shutter speeds if it makes the post workflow better.
did you use pro version of real smart motion blur or just regular?
What's the framerate of your input footage?
I couldn't see it until you showed it in slow motion.
Very useful video! It was easy to spot the real one, but mostly because I tested this myself and found digital motion blur worked really bad for flying low and get that sense of speed. Do you know of any high end software that can do it better?
You shouldn't need digital motion blur in low light scenarios if your shutter speed is already going to be the slowest you can push it. If your out in the daytime you might want to add motion blur because otherwise you would need an nd filter .
Awesome video, thanks for the comparison, exactly what I was looking for 🙏
i thought the top right looked best, the real motion blur with that judder from realsteady is a deal breaker for me. i can't use AE, my computer cries when i export normal movies. i'm sure i seen a shutterspeed ''control'' in post somewhere.... might have been old imovie. i do understand that motion blur is cool and sort of directs the eye through the flight but i'm not a big fan of it. seems to me if i want to glance at a detail i should expect it to be as crisp as anything else. ps. i think the swirling effect on the shrubs was mainly the wind?
I think modern eyes are simply more used to high shutter speeds. Watching tons of UA-cam and less Hollywood etc. Sometimes it can ruin a shot for me (high shutter speed) and make it look jarring and stuttered.
Other times I think the opposite. And think the slow shutter makes it look smeary and you lose the crisp feeling.
So it’s almost different for every shot. :/
LOL, I could tell the bottom left was different from the others so I thought it was the fake. I was totally wrong!
Instant sub. Great effort. Thank you.
Will fly smoother and keep the ND on for now. Hopefully the tech will get better.
While there are some advance techniques one can do in Fusion or Nuke with optical flow and combination with other nodes and tricks, most of the time its not worth it, because it takes long to render out. By the way that is also something one can use to fix rolling shutter and other problems, like bad frames etc. Optical flow in Fusion combined with other nodes is quite versatile, but again takes too long to render for most types of footage. Best shoot real thing.
What is a high end motion blur plugin?
'Degrading' and image- essentially what motion blur is doing- is relatively easy for software/AI to do- so the artificial blurs should be (and are) very good... (Sharpening/up-rezing in post on the other hand is much harder- trying to create new pixels where there were none- just a guess...)
how about filming in 60fps and use the I-frames for information for motion blur when converting to eg 30p.
This is a great question, more data should be able to help the algorithms create more realistic motion blur
Right at the beginning of the video, when the guy walks in, the fake ones show a weird blur around the guy's legs.
Why some artifacts and some shakiness appears when using GoPro 8 and ReelSteady? I mean, there are some frames that the image gets difused, like out of focus. Is this related to vibrations? How to get crystal clear images with ND Filters, GoPro 8 and ReelSteady? Any specific settings for that? Thanks!
I talk about this in the vid. Yes as I said in the vid it was very windy that day. And gusts make the quad twitch a bit. Which are baked into the footage when you use slowish shutter (I was at 180deg for the ND shot)
This is the downside to doing it in camera.
My quads and props are very smooth and I can normally shoot at 180deg no problem if there is no wind. (But I do live by the sea so it is often windy, in which case I speed up the shutter speed to often 1/100 and sometimes 1/200.
Right side where easy to spot. I was wrong with the one on the left side! Ps: Does anyone know a tutorial on how to do this in Kdenlive or Shotcut?
My money went on Top Right. 😪
Are you using RE:Vision FX RSMB?
Details are in description.
Someone else rendered fake blur
@@biscuitsalive Thank you!
Bottom Left is my guess
I spotted it in the bottom left when the copter was hovering and the background trees were moving in the wind. Only the bottom left showed the tree blurry. that was the give away
I spotted it on literally the first few seconds as soon as the guys foot entered the scene. Top left blurs the grass on the bottom of the foot so thats not it, top right same thing but now ontop of the foot, and bottom right same issue with top left but not as intense.
You can also tell by the distortion on certain frames. Some of the motion blur will abruptly end causing a clear border like the bottom right but top left and right exhibit way harsher distortions, particularly on the jeans.
Not gonna lie, that jitter on 1/50 is more annoying than the swirls on the fake motion blur.
i personally spotted it easily after the clap at the beginning
With this darn UA-cam compression , all are artificially blurred. lol I hate UA-cam compression.
You have to watch in 1440 or 4k.
Not too bad then. 1080 is terrible.
@@biscuitsalive Right. Downloading it now at highest resolution to take a look offline. But later when you start explaining it and pause it quite obvious what the difference is. Still worth checking out. Thanks for the video.
Ok I'm playing fair... took me about 45 sec to spot the bottom left is real because it has that horrible blur the whole frame effect when you use post stabilization, not possible to recover a blurry source image. Did I win? only future me will know...
Yay! - also would be interested to see how the fake motion blur is handled with different frame rates like 60 vs 120 etc. More frame data presumably would yield a better result?
Just check frame by frame 00:01 first we see he is stepping into the frame and there you can tell which is fake which is real. But funny thing is bottom right corner is the best motion blur in terms of realism and it takes 6 mins to render lol.
frame by frame hotkeys are " , " and " . "
top left fake. bottom left real. occasional bad vectors in the fake one. occasional blurry frames from shake in the real.
Bottom left isn’t the same as the others !! I’m guessing that that’s real motion blur . Bottom left
Holy crap. I got it.
good spot, i dont think many people would get it, maybe viewing on a big screen in 4k its much more obvious, but not on a phone
@@biscuitsalive generally I’d agree but I watched it on an iPhone X.
Honestly though. If the footage wasn’t side by side and I couldn’t pause it to have a look at each frame then I wouldn’t have known it was post production.
@@erikdeeNOSPELLSNO lol
@@erikdeeNOSPELLSNO if you want And @biscuitalive is willing I’d like him to try it again without the results.
Honestly I didn’t cheat. Didn’t even know the results would be in this video.
who can tell with UA-cam compression and billions of leaves in the video?
Lower left is real
i've got seasick from this
Sorry to hear that. :)